Reflection angle domain (RAD) imaging is gaining interest as an alternative to shot and offset domain methods Xu et al. (1998). The advantage of RAD imaging is that it reduces the number and strength of artifacts seen in complex areas, particularly artifacts caused by multipathing. RAD imaging reduces artifacts but does not overcome the inherent limitations in surface seismic recording geometries. As a result, areas where the RAD is most useful can also benefit from replacing migration with inversion. Although inversion is more expensive from a computational standpoint, it can effectively address amplitude problems Duquet and Marfurt (1999) and, with intelligent preconditioning, null space concerns caused by limited survey geometry. In this paper, we use a preconditioned inversion approach described in Prucha et al. (1999b). By using preconditioning, we are attempting to fill in model components that have no corresponding data. This raises the concern of whether or not the preconditioning is creating a reasonable and realistic model, especially in terms of amplitude.
Amplitude analysis is difficult in complex areas, even when we have an accurate velocity model Castagna and Backus (1993). Limited recording geometry and shadow zones caused by overburden can create artificial, erroneous amplitude variation with angle (AVA) results. Since our preconditioned inversion helps fill this model null space, we would like our created amplitudes to be as reasonable as possible. In this paper, we will show that our chosen preconditioning operator fills the model null space intelligently with amplitudes that are more accurate than those obtained by migration alone.
To show the benefits of preconditioning in the RAD, we will first expand on why it is necessary to use an inversion process. We will then explain how we carried out our preconditioned inversion. Finally, we will apply our method to a synthetic model and show that the amplitude response after RAD preconditioning is more accurate than that after simple RAD migration.